List your company
ResearchFormerly WizWand

Accuracy on MMLU-Pro

82.05AccuracySMCS
-1.274820.357641.9963.6224Jul 14, 2025Aug 29, 2025Oct 15, 2025Nov 30, 2025Jan 16, 2026Mar 3, 2026Apr 19, 2026
Updated 5mo ago

Evaluation results

MethodLinks
SMCS2025.0782.05
GPT-4.12025.0780.41
QwQ-32B2025.0774.69
Self-MoA2025.0769.76
HuggingGPTImplementation Framework=original2026.0365.59
AgentVerseImplementation Framework=original2026.0364.64
AgentVerseImplementation Framework=MASFactory2026.0364.16
HuggingGPTImplementation Framework=MASFactory2026.0363.66
CAMELImplementation Framework=MASFactory2026.0363.04
OpenReasoner-Zero*Method Category=Previous RLVR Methods, Result Source=Reported from original papers2025.1058.7
DIVERBase Model=DeepSeek-R1-Distill-Qwen-7B2025.0956.6
GRPOBase Model=Qwen2.5-7B-Base, Reward Strategy=Clip-higher2025.0955.4
DIVERBase Model=Qwen2.5-7B-Base2025.0955.2
GRPOBase Model=DeepSeek-R1-Distill-Qwen-7B, Reward Strategy=Clip-higher2025.0954.9
LUFFYMethod Category=Comparable Baseline Methods2025.1053
Prefix-RFTMethod Category=Comparable Baseline Methods2025.1052.1
DIVERBase Model=LLaMA-3.1-8B-Instruct2025.0952
ReLIFTMethod Category=Comparable Baseline Methods2025.1051.9
Vibe Graphing-Task SpecificImplementation Framework=Vibe Graphing2026.0351.73
GRPOBase Model=LLaMA-3.1-8B-Instruct, Reward Strategy=Clip-higher2025.0950.8
CAMELImplementation Framework=original2026.0350.08
SFT+RLMethod Category=SFT and RL2025.1049.6
Dream-Instruct 7B + G-StarRefinement Strategy=Loop-based refinement2025.1047.9
ICPO†Method Category=Comparable Baseline Methods2025.1047.6
Dream-Instruct 7BSource=Reproduced2025.1046.9
ICPOMethod Category=Comparable Baseline Methods2025.1046.5
SFTMethod Category=SFT and RL2025.1044.4
Dream-Instruct 7BSource=Originally Published2025.1043.3
Falcon3-7BEvaluation Protocol=Single, Backbone Model=Falcon3-7B2025.1042.71
RLMethod Category=SFT and RL2025.1042.6
M_RLP + PostPre-training Condition=Reinforcement Learning from Pretraining, Post-training=SFT + RLVR, Evaluation Protocol=Standard2025.0942.4
K-LVR (MCV)Evaluation Protocol=Ensemble (PoE), Vocabulary Strategy=MCV2025.1042
Oat-Zero*Method Category=Previous RLVR Methods, Result Source=Reported from original papers2025.1041.7
M_RLP + PostPre-training Condition=Reinforcement Learning from Pretraining, Post-training=SFT + RLVR, Evaluation Protocol=@1[4]2025.0941.3
K-LVR (1-Bytes)Evaluation Protocol=Ensemble (PoE), Vocabulary Strategy=1-Bytes2025.1041.21
ReConcileBase Model=Gemma-3-4B2025.0840
M_CPT + PostPre-training Condition=Continuous Pretraining, Post-training=SFT + RLVR, Evaluation Protocol=Standard2025.0939.92
M_CPT + PostPre-training Condition=Continuous Pretraining, Post-training=SFT + RLVR, Evaluation Protocol=@1[4]2025.0938.49
M_base + PostPre-training Condition=Base, Post-training=SFT + RLVR, Evaluation Protocol=Standard2025.0937.85
Qwen2.5-3BEvaluation Protocol=Single, Backbone Model=Qwen2.5-3B2025.1037.71
M_base + PostPre-training Condition=Base, Post-training=SFT + RLVR, Evaluation Protocol=@1[4]2025.0936.53
M_RLPPre-training Condition=Reinforcement Learning from Pretraining, Post-training=None, Evaluation Protocol=Standard2025.0934.62
SimpleRL-Zero*Method Category=Previous RLVR Methods, Result Source=Reported from original papers2025.1034.5
PRIME-Zero*Method Category=Previous RLVR Methods, Result Source=Reported from original papers2025.1032.7
Single-Agent BestBase Model=Gemma-3-4B2025.0832.3
DIVERBase Model=Qwen2.5-Math-1.5B2025.0931.8
TMA-AllComponBase Model=Gemma-3-4B2025.0831.7
M_RLPPre-training Condition=Reinforcement Learning from Pretraining, Post-training=None, Evaluation Protocol=@1[4]2025.0930.8
MDAgentsBase Model=Gemma-3-4B2025.0830.7
GRPOBase Model=Qwen2.5-Math-1.5B, Reward Strategy=Clip-higher2025.0930.2
DyLanBase Model=Gemma-3-4B2025.0828.7
M_basePre-training Condition=Base, Post-training=None, Evaluation Protocol=Standard2025.0928.17
M_CPTPre-training Condition=Continuous Pretraining, Post-training=None, Evaluation Protocol=Standard2025.0927.81
MedAgentsBase Model=Gemma-3-4B2025.0824.7
M_CPTPre-training Condition=Continuous Pretraining, Post-training=None, Evaluation Protocol=@1[4]2025.0924.61
M_basePre-training Condition=Base, Post-training=None, Evaluation Protocol=@1[4]2025.0923.95
COACTTraining Dataset=MATH2026.0423.42
Pref + EntTraining Dataset=MATH2026.0422.76
EntropyTraining Dataset=MATH2026.0422.15
Pref CertaintyTraining Dataset=MATH2026.0422.08
RandomTraining Dataset=MATH2026.0421.34
Qwen2.5-Math-7BMethod Category=Base Model2025.1016.9
Naive (MCV)Evaluation Protocol=Ensemble (PoE), Vocabulary Strategy=Naive (MCV)2025.102.07
UnionEvaluation Protocol=Ensemble (PoE), Vocabulary Strategy=Union2025.101.93